With conventional coding techniques using blocks, a neighbouring block is defined on the basis of self correlation of an image and the encoding operation is conducted on a policy of eliminating redundancy within the block. Various shapes are found when evaluating an image by means of neighbouring blocks, including those for plane areas and those for edges.
In ordinary encoding operations, while redundancy may be eliminated with ease from plane areas in order to compress the data to a large extent, edges are disadvantageous in terms of compression because of the large amount of information involved there so that distorted and unbalanced quantization can occur as a result of encoding. While the resultant degradation of the image quality may be unrecognizable in certain occasions due to the masking effect of the sight, an increased amount of information has to be used to alleviate the degree of visual degradation.